[AFRINIC-rpd] Academic IPv4 Allocation Policy Second Draft (AFPUB-2013-GEN-001-DRAFT-02)

[Owen DeLong]

On Jan 28, 2013, at 05:01 , Sunday Folayan <sfolayan at gmail.com> wrote:

> On 28/01/2013 10:06, Nii Narku Quaynor wrote:
>> Thanks. What prevents a university from receiving a 10:1 via current policies
> 
> Justification, a la Network diagram!
> 
> Recall the thread from Andrew that started this all! https://lists.afrinic.net/pipermail/rpd/2012/002553.html
> 
> It is easy to justify 48 users connected to a 48 port switch, by proving/justifying the fact that you actually bought 1000 switches, so you can connect 48,000 or less users.
> 
> How do you do that with Wireless Access points? How many users per AP? etc? Not that easy like fixed infrastructure. This was the frustration with coming up with an acceptable justification which Andrew expressed and along the thread, which led to my challenging him that we fix it, instead of complaining.
> 

Actually, anyone who has experience doing WiFi for meetings and conferences with a high geek factor
can pretty quickly tell you that anything > about 50 users per AP tends to degrade pretty rapidly.
Anything north of 100 users on 2.4Ghz is virtually unusable and anything north of about 200
is virtually unusable at 5Ghz.

That is why if you go to any networking conference in most places, instead of seeing a few high power
access points to cover the large area of the conference, you will see many low-power APs spread
around. For example, to cover the main hall of a conference with ~500 attendees, you'll usually want
somewhere between 12 and 20 APs in the room.

However, with transient users, this becomes more problematic because your DHCP pool may have
a lease time of several hours where users may only be active for ~30 minutes at a time. In such a
case, the addresses are tied down for much longer than the AP channel space is in use and this
can have a multiplicative effect on the address demand over the available channel space.

This can be equally true for wired ports serving transient users in a library or other location offering
network jacks for public transient use. As such, a 48 port switch may actually support a requirement
for many more than 48 users. Similarly, an AP may also have a need to support a larger number.

OTOH, there is usually a loose correlation at a university between expansion in the student population
and expansion in other infrastructure and in the size of the staff. As such, while not perfect, the
number of students to be supported multiplied by an appropriate ratio (I suggest 5 as a reasonable
default given what has already been discussed) does actually work reasonably well as a rule
of thumb or yardstick by which to easily judge a reasonable amount of address space for said
institution.

I think it is in the best interests of the region to remove barriers to deploying addresses to
educational institutions. I'm not sure I agree with the idea of limiting the policy to higher
education institutions, but I'm not sure whether it will matter in the remaining time available
for IPv4.

Owen

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.afrinic.net/pipermail/rpd/attachments/20130128/c11a375a/attachment.html>